Looking through walls - Coatings on glass for ... - W. Theiss Hard

<strong>Looking</strong> <strong>through</strong> <strong>walls</strong> - <strong>Coatings</strong> on glass for buildings
by Wolfgang Theiss
M. Theiss – Hard- and Software for Optical Spectroscopy, Dr.-Bernhard-
Klein-Str. 110, D-52078 Aachen, Germany
Abstract
Modern buildings are equipped with windows having thermal insulating
properties like solid <strong>walls</strong>. Advanced thin film coatings not only efficiently
suppress the emission of heat radiation but provide a large variety of possible
colors for architectural design. Optical spectroscopy from the ultraviolet to the
infrared plays an important role in the development of these products. The
relation between deposition conditions and optical properties of the produced
materials is studied, and the interaction of the individual layers with their
neighbours in the stack is investigated. The optical design of optimized
multilayers and spectroscopic production control is discussed.
Moderne Gebäude sind mit Fenstern ausgestattet, die thermisch so gut
isolieren wie massive Wände. Die dabei eingesetzten Dünnschichtsysteme
unterdrücken effizient die Wärmeabstrahlung des Glases und geben
gleichzeitig dem Architekten eine grosse Gestaltungsmöglichkeit bezüglich der
Farbe der Verglasung. Optische Spektroskopie vom ultravioletten bis zum
infraroten Spektralbereich spielt eine große Rolle bei der Entwicklung dieser
Produkte. Der Zusammenhang zwischen den Depositionsbedingungen und den
erzielten optischen Eigenschaften der produzierten Materialien sowie die
Wechselwirkung der einzelnen Schichten mit ihren Nachbarn im
Schichtsystem werden untersucht. Das optische Design der
Vielfachschichtsysteme und die spektroskopische Produktionskontrolle werden
diskutiert.
Introduction
The way we build houses and large buildings has changed dramatically during the last
century. A comparison of two modern office buildings (see figs.1 and 2) shows that
solid <strong>walls</strong>, giving shelter from wind and rain as well as low and high outside
temperatures, are more and more replaced by transparent windows.
This change in architecture is based on two industrial developments, the large scale
production of almost perfectly flat glass and the large area coating technology. The
mass production of glass panes is possible using the float process where a continuous
stream of glass moves (on a bath of liquid tin) from the hot liquid state to the solid
state. Pieces of typically 6 m * 3.20 m are cut at the end of the line. The pane
production is followed by a coating step: The thermal properties of uncoated glass
would lead to a tremendous heat loss if the temperature inside an office building like
the one shown in fig.2 is significantly higher than outside. On sunny days in the
summer, on the other hand, buildings equipped with uncoated glass would strongly
heat up and produce significant costs for air-conditioning. Thin film systems on glass
avoiding heat losses are called low-e coatings, whereas so-called 'solar control

coatings' avoid to some extent
the building's heating up by
sunlight. The function of both
coating types is discussed in the
next sections.
The cost-effective production of
large area coatings is a
permanent struggle for
homogeneity: The deposition of
layers with thickness variations
of less than 1 nm on a 6 m * 3.20
m substrate would be difficult
enough [1]. Unfortunately, this
problem is doubled by the fact
that the sputtering devices must
be operated in a way causing their characteristics to drift away in time. The section
about thin films deposited by
large area coaters addresses
the problems related to this
deposition technique. In
addition, the interaction of
adjacent layers during and
after the deposition is
discussed.
The design of new coating
products must take into
account the wanted target
properties of the coating, the
available range of optical
constants and film
thicknesses in the production
line, possible homogeneity
problems and costs.
The last section is about the
optical inspection of the
Fig. 2 A large modern office building in the year 2004
production which is an
important tool to control the
deposition. Information about
thicknesses and optical constants of the layers produced at present is used to take
appropriate deposition control actions to obtain wanted and stable coating properties.
All computations shown in this article have been done with the CODE software [2].
Low emission coatings
Fig. 1 A large modern (700 years ago) office building
Replacing a solid wall by a piece of glass does not only increase the <strong>through</strong>put of
light but also that of heat. The transfer of heat <strong>through</strong> a wall is usually specified by
the so-called U-value (overall coefficient of heat transmission) [3]. This quantity gives
the transmitted power per wall area and temperature difference between the inside and
the outside of the building. The usual unit is W/(m 2 K).
A 23 cm thick wall of simple brick stone has a U-value of 2.2 W/(m 2 K) which means
that there is a power stream of 22 W <strong>through</strong> each square meter if the inside-outside

temperature difference is 10°C. Cavities and additional heat insulation (e.g. by foams)
can reduce the U-value of a wall to 0.5 W/(m 2 K) or less.
An uncoated glass pane has a U-value of about 6 W/(m 2 K) which is 10 times higher
than that of a typical wall. Windows with single glazing can easily reach temperatures
below 0°C in cold winter nights. While this may lead to beautiful ice sculptures
growing in the kitchen there is a strong demand to decrease the U-value and avoid
energy streaming out of the window.
The easiest way to block the heat conduction <strong>through</strong> the glass is to introduce an air
gap (typical thickness: 10 ... 20 mm) between two glass panes: Windows with double
glazing have reduced U-values of about 2.7 W/(m 2 K). The remaining heat transfer
mechanisms in double glazings are heat convection <strong>through</strong> the air gap and the
trransfer of energy by infrared radiation. The convection losses can be reduced
replacing the air by 'heavy' filling gases (Ar is standard today, Kr and Xe would be
better but are more expensive). However, the U-value is lowered only by about 0.1 ...
0.2 this way.
In order to
understand the
energy transfer by
infrared radiation
we have to
investigate the
properties of float
glass for
wavelengths above
5000 nm (far
infrared). Here the
transmission of a
typical 4 mm pane
is zero, and the
reflectance is a few
percent only (see
fig.3). This means
that glass is almost completely 'black' for mid and far infrared light. A material that
absorbs most of the incident radiation is, on the other hand, also a very efficient
U-value
3.0
2.8
2.6
2.4
2.2
2.0
1.8
1.6
1.4
1.2
1.0
R / T / Intensity [a.u.]
100
80
60
40
20
0
Blackbody radiation and float glass properties
Transmittance
Reflectance
300 K
290 K
5000 10000 15000 20000
Wavelength [nm]
Fig. 3 Transmittance and reflectance of a 4 mm float glass from the UV to
the far infrared. In addition, the spectral distribution of blackbody radiation
emitted at 290 and 300 K is shown for comparison.
U-value vs. Ag thickness
Outside
Inside
Vacuum
Float A
4.000 mm
Argon
16.000 mm
Ag
12.0 nm
Float A
4.000 mm
Vacuum
0 5 10 15 20 25 30 35 40
Ag thickness [nm]
Fig. 4 U-value of a double glazing with a silver coating on the
interior pane to suppress infrared emission. The window has a
16 mm Argon filling.
emitter of radiation. Hence a
glass pane acts as an almost
perfect blackbody radiator.
In a double glazing window
one has two efficient infrared
light sources and absorbers
opposite to each other. If
there is a temperature
difference between the
interior of the building and
the outside, a quite large net
flow of radiative energy from
the warm side to the cold
occurs. Fig. 3 shows the
spectral distributions of
blackbody radiation emitted

at 290 and 300 K.
In order to suppress the
radiation exchange it would
be nice to be able to switch
off the 'warm' light source.
This can be achieved by the
deposition of a metallic layer
on the pane mounted on the
interior side of the building.
The layer increases the
reflectance, decreases the
absorption and this way
diminishes the emission of
infrared radiation. Silver
turned out to be the best
material for this purpose. Fig.
4 shows the strong effect of the silver thickness on the U-value. U-values significantly
below 1.0 W/(m 2 K) cannot be achieved with double glazings. With triple glazings
values of 0.6 W/(m 2 K) are possible.
However, the positive improvement of the heat insulation is accompanied by the
unwanted side effect of severely decreasing transmittance in the visible (Fig. 5).
Fortunately, surrounding the silver layer by two dielectric layers (oxides in most
cases) of appropriate
Transmittance
1.0
0.8
0.6
0.4
0.2
0.0
thickness and refractive
index one can almost
completely restore the
transmittance of the
uncoated glass while the
suppression of the infrared
emission remains
unchanged. Thin film
systems like this are
called low-e (for 'low
emission') coatings (see
fig. 6).
The oxide layers of low-e
coatings can be used to
vary the color of the
window. Selecting different types of oxides and different thicknesses for the
individual layers one can adjust the visual appearance of a building to the architect's
ideas (see fig. 7).
Solar control coatings
1.0
0.8
0.6
0.4
0.2
0.0
400
450
500
550
Wavelength
[nm]
600
650
700
750
Transmittance
0 5 10
0.0
15 20 25 30 35 40
Ag thickness
[nm]
Fig. 5 Dependence of the transmittance of a double glazing
(layer structure as in fig.4 ) on the silver thickness.
Uncoated double glazing
10 nm silver coating
Low-E coating
400 450 500 550 600 650 700 750
Wavelength [nm]
Fig. 6 Transmittance spectra of an uncoated double glazing, a
double glazing with a 10 nm silver layer on the interior pane, and a
typical low-e coating (also with 10 nm silver)
If a large fraction of a building fassade is made of glass a lot of sunlight is transmitted
into the building. Most of it will be absorbed inside and significantly heat up the
rooms on sunny days. This can cause high air-conditioning costs. Whereas the
transmission of light in the visible cannot be avoided (that's the purpose of the
window!) the large fraction of infrared light in the solar radiation arriving at a
building (shown in fig. 8) is unwanted and can be blocked by so-called 'solar control'
coatings.
1.0
0.8
0.6
0.4
0.2

Transmittance / Intensity
Fig. 7 Color variation of low-e coatings with the same U-value of 1.2 W/(m^2 K)
1.0
0.8
0.6
0.4
0.2
0.0
500 1000 1500 2000 2500
Wavelength [nm]
Fig. 8 Transmission spectra of double glazing windows without coating, with low-e and solar control
coatings. The intensity distribution of incident solar radiation (the so-called AM 1.5 radiation
distribution is taken from [3]) is shown as well.
Fig. 8 shows a comparison of the transmittance spectra of typical low-e and solar
control coatings. The latter is much more 'selective', i.e. there is a sharp decrease of
the transmission from the visible to the infrared. This ideally step-like feature cannot
be obtained with a single silver layer. In most cases, two silver layers being embedded
in and separated by oxide layers are applied to achieve high selectivity. Solar control
coatings are deposited on the inner side of the exterior pane in double glazing
windows. Their performance is expressed in two quantities: The 'Light transmittance'
(average transmittance in the visible, weighted with our eyes' sensitivity) [5] indicates
how well we can look <strong>through</strong> the window. The 'Total solar energy transmittance' g
[5] gives the overall <strong>through</strong>put of solar radiation, including the direct solar
transmittance and secondary energy streams by absorption and re-emission. A good
solar control coating has a high light transmittance and a low g-value. A ratio of 2
between the two quantities is considered to be almost ideal.
<
>
Uncoated double glazing
Low-E coating
Solar control coating
AM 1.5 radiation
<
>

Thin films produced by large area coaters
The layer stacks discussed above must be deposited on large glass panes with a very
high degree of homogeneity and at reasonable costs. At present the production scheme
sketched in fig.9 is the method of choice [6]. The uncoated panes are moved into a
large evacuated volume. Linear sputtering devices (usually operated with Ar gas)
Uncoated glass
Pumps
Fig. 9 Principle of a large area coating line for the production of a solar control coating
optimized for homogeneity perpendicular to the motion of the glass deposit metals or
(with additional reactive gas inlets) oxides or nitrides. Depending on the speed of the
substrates, the required layer thickness and the available sputtering rates several
cathodes are grouped together to produce one of the coating's functional layers.
To maintain the vacuum and to properly separate the 'oxide areas' from the 'metal
areas' a tremendous pumping equipment is required. The sputtering devices consume a
lot of electrical power, too.
Hence the goal is to finish
as many panes as possible
without interruption –
typically every minute a
coated pane leaves the
coater. If everything works
fine about 25000 m 2 glass
can be coated on one day.
However, reactive
sputtering of oxides is a
stable process only in the
so-called oxidic mode
where the partial pressure
of oxygen is very high.
Unfortunately, in this mode
the deposition rates are
very low (see fig. 10).
Large area coating line
Sputtering device Vacuum Coated glass
Bottom Oxide Metal Center Oxide Metal Top Oxide
Deposition rate [a.u.]
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
Metallic mode
Transition area
Oxidic mode
0 2 4 6 8 10
Oxygen flow [a.u.]
Fig. 10 Typical relation between the sputtering rate and the
applied oxygen flow. In the transition area the deposition rate for a
given flow depends on the history of the process (hysteresis
behaviour).
Economic coating deposition is only possible in the transition area where the
stabilization of the sputtering process is very difficult [7]. This problem and the slow
drift of the sputtering characteristics due to erosion of the targets (due to the loss of
material which is used for the deposition) make it necessary to constantly watch and
regulate the process.

For the proper design of coatings it is very important to investigate the dependence of
the obtained optical constants on the deposition conditions, e.g. the applied electrical
power and oxygen (or nitrogen) pressure. The best way to do this is to produce single
layers of all relevant materials and analyze optical spectra (like reflectance and
transmittance) to determine the complex refractive index n + i k of the sputtered
materials. With appropriate adjustable dispersion models (like the OJL model for
amorphous materials [8]) one can describe the optical properties of all materials used
in glass coatings with a few key parameters (see fig.11). The relation of these
parameters to the deposition conditions gives an excellent basis for coating design and
production control. This increased knowledge justifies to use the expensive
production line for several hours to produce single layers under various production
conditions.
2.0
1.5
1.0
0.5
0.0
Air
Oxide
100.9 nm
Float glass
4.000 mm
Air
20000.0000
Bandgap [1/cm]
31463.4570 50000.0000
Thickness [nm]
100.9
Refractive index Oxide analysis (single layer)
0.0000
Constant
3.1439 5.0000
500 1000 1500 2000 2500
Wavelength [nm]
R (coating side)
500 1000 1500 2000 2500
Wavelength [nm]
Fig. 11 Analysis of three measured spectra (the red spectra on the right) by fitting a dispersion model
and the thickness of an amorphous oxide layer on glass. The simulated spectra are drawn blue. The
graph of the obtained refractive index model (left side) shows the real part n in blue and the imaginary
part k in red.
Knowing single layer properties is not sufficient to describe spectra of complex
coating products with high quality. Layers in a layer stack can have different
properties than single layers on glass. The growth of a thin film depends on the type,
temperature and roughness of the underlying material. In addition, high rate sputtering
is not a very gentle method, and the deposition of a layer may influence (damage) the
films underneath. Slow diffusion processes may lead to changes of the coatings even a
long time after finishing the deposition.
The silver layers which are very important for the function of low-e and solar control
coatings are very sensitive to their neighbourhood in the layer stack. The high
reflectance of thin silver films depends critically on the mobility of the electrons.
Their free motion can be disturbed by impurities and defects in the bulk, but also by
surface roughness. In a metallic film of a few nanometers thickness there are frequent
boundary collisions of the electrons – smooth interfaces lead to mirror-like reflections
of the electron waves with no decrease of the mobility whereas rough surfaces cause
diffuse electron scattering with significant consequences for the conductivity. The
damping constant of the simple Drude model for the electrons [9] in silver is a good
parameter to characterize the quality of the layer – fig.12 shows the influence of the
Modified Transmittance
Modified Reflectance
Modified Reflectance
40
35
30
25
20
15
10
5
0
100
90
80
70
60
50
40
30
20
10
0
40
35
30
25
20
15
10
5
0
T
500 1000 1500 2000 2500
Wavelength [nm]
R (glass side)
500 1000 1500 2000 2500
Wavelength [nm]

silver quality on the U-value of the final product (low-e coating).
Achieving and maintaining a good silver conductivity can be very important in order
to place a coating product on the market. In order to reach this goal, one has to do
extra work on both sides of the silver layer. Underneath surface roughness is reduced
U-value
U-value vs. electron damping
1.50
1.45
1.40
1.35
1.30
1.25
1.20
1.15
1.10
1.05
1.00
200 300 400 500 600 700 800 900 1000
Damping constant [1/cm]
Fig. 12 Relation of the U-value of a low-e coating to the damping constant of the electrons in the
silver layer (Drude model). For sputtered layers, typical values for the damping constants are 300 to
500 1/cm. 'High quality silver' can have damping constants as low as 140 1/cm.
by the deposition of thin oxide layers specialized to build very smooth interfaces [10].
On top of the silver an additional blocker layer is produced which protects the silver
from subsequent sputtering damage and diffusion of oxygen atoms. In some cases, the
final coating is mechanically protected by an extra-hard nitride topping. This way lowe
coatings which have in principle the structure oxide / silver /oxide can easily be
composed of 6 to 8 thin films.
Window coating design
Outside
Inside
Vacuum
Float A
4.000 mm
Argon
16.000 mm
B-oxide
56.3 nm
Ag
10.1 nm
C-oxide
23.5 nm
Float A
4.000 mm
Vacuum
The design of new coating products should be based on the knowledge about
achieveable n and k values, interactions of neighboured layers and established
deposition rates. The quantitative analysis and description of single layer samples and
already existing coating products (see fig. 13) can be used to build up a database of
optical constants and partial layer stacks. This can be used as a source of successful
building blocks for new combinations.
Since the operation of large area coating lines is very expensive one can save a lot of
money if the prediction of the optical properties of new products is reliable. Ideally
the parameters of the model should reflect the parameters of the production devices
such as electrical power or oxygen pressure. The design software should enable easy
selection and variation of these parameters with instant display of the results, i.e.
optical spectra and characterizing quantities such as color coordinates, light
transmittance and U- or g-values. In addition, it should be easy to inspect how
production tolerances (e.g. pressure fluctuations) show up in the properties of a
coating.
After manual parameter variations and visual inspection one would like to switch to
automated design. Target spectra or wanted values for colors or other technical data

are used as optimization goals. Usually several parameters related to the thicknesses
and optical constants of the model are adjusted simultaneously. For parameter fine-
Reflectance
Transmittance
1.00
0.80
0.60
0.40
0.20
0.00
1.0
0.8
0.6
0.4
0.2
0.0
Solar control coating: Quantitative modelling
R
500 1000 1500 2000 2500
Wavelength [nm]
T
500 1000 1500 2000 2500
Wavelength [nm]
1.00
0.80
0.60
0.40
0.20
0.00
Fig. 13 Example for successful quantitative modelling of a solar control coating with 14 layers. The
measured (red) reflectance from the coating side (R), the transmission (T) and the reflectance with
backside illumination are reproduced very well by the model (simulated spectra in blue). There are no
measured data for the frontside reflectance at an angle of incidence of 70°.
tuning one can use optimization methods that move from the starting design to the
next local minimum of the deviation to the target values. This is a matter of seconds
or minutes and can be mixed with manual user interactions. Methods that search for
the global deviation minimum are much slower and should not require any user
actions in order to be applied in overnight optimization runs.
Optical production control
Reflectance
Reflectance
1.0
0.8
0.6
0.4
0.2
0.0
As discussed above, thin film deposition in large area coating lines must be permantly
observed and corrected. Due to the high operational costs, one must detect deviations
from the target properties of the coating as early as possible.
Fig. 14 shows the principle of an optical production control system [6]. At every
important position along the production line appropriate optical measurements are
done. An 'optical network' is responsible for the coordination of the data acquisition
and analysis of the measured data. It provides status information about the present
condition of the production line. The operator can use this information in order to
decide if production parameters should be changed and which actions are required.
It would be advantageous to record several spectra at each position: Reflectance from
the coating and the glass side as well as transmittance, if possible in a large spectral
range. These spectra would provide enough information to safely determine the
thickness and the optical constants of each main layer of the coating. Ideally one
would like to measure at several spots (at least 3) perpendicular to the direction of the
glass motion in order to check the lateral homogeneity of the deposition.
There are some limitations, however. Recording 3 spectra at 5 positions along the
production line with 3 spots perpendicular to the line would mean to buy, install and
operate 45 spectrometers. Even if low-cost array spectrometers are used, this is just
too expensive at the moment. Also the analysis and the handling of the obtained data
R back
500 1000 1500 2000 2500
Wavelength [nm]
R, 70 deg
500 1000 1500 2000 2500
Wavelength [nm]
Vacuum
Material 1
5.0 nm
Material 7
21.4 nm
Material 6
3.0 nm
Material 4
10.9 nm
Material 5
1.2 nm
Material 3
4.4 nm
Material 2
8.4 nm
Material 1
10.0 nm
Material 7
35.4 nm
Material 6
3.0 nm
Material 4
12.2 nm
Material 5
1.2 nm
Material 3
4.4 nm
Material 2
13.8 nm
Float
7.900 mm
Vacuum

would require several computers and a sophisticated software which would also be
expensive.
A good current compromise is to record transmittance spectra in the coating line,
Uncoated glass
Pumps
Fig. 14 Scheme of an optical production control system. After the deposition of each functional layer
optical measurements are taken (vertical yellow bars). The 'Optical inspection network' is responsible
for proper coordination and analysis of the measurements and delivers an overview of the present
deposition results.
only in the center of the panes. The final product is inspected outside the vacuum
chamber with a moving system of 2 or 3 spectrometers (R, T, eventually R from the
backside) which record spectra at several spots perpendicular to the glass motion (see
fig. 15). This way about 10
spectrometers are required.
Another restriction concerns the
available spectral range. Since
there is not too much time for
spectrum recording, array
detectors have to be used. These
are available at low cost (USD
2000 ... 4000) only in the
Vis/NIR range up to 1100 nm
wavelength (since they use
silicon detectors). Infrared array
detectors which extend the range
up to 1700 nm or 2200 nm
wavelength are about a factor 10
more expensive (USD 20000)
and up to now not widely used
for online deposition control.
With increasing complexity of
the coatings optical
Large area coating line
Sputtering device Vacuum Coated glass
Bottom Oxide Metal Center Oxide Metal Top Oxide
Optical inspection network
Fig. 15 Possible spectrometer positions for production
control:
Left: Measurements only at the center of the deposition
Center: Measurements at several fixed positions
Right: Moving spectrometer system
measurements providing information about the individual parts of the deposition line
will become more and more important. Not only because information about individual
layers cannot be extracted from the final coating properties any more if the number of
layers is too large, but also because small deposition errors in the beginning of the
production line can eventually be compensated by proper parameter changes at

subsequent deposition steps.
Summary and outlook
Progress in glass coating technology has enabled architects to replace solid <strong>walls</strong> by
transparent glass. Complex large area coating lines produce very homogeneous thin
film systems with tailored optical and thermal properties at low costs. Optical
spectroscopy is used in research and production control to determine layer thicknesses
and optical constants.
Currently under development are switchable optical properties (electrochromic or
gasochromic) and the integration of large area displays and thin film solar cells into
glass fassades.
References
[1] M.Geisler et al., Proc. 4 th ICCG (2002), p. 59 , C.P. Klages, H. Gläser, M.A.
Aegerter (eds.)
[2] CODE – a thin film analysis and design program developed and distributed by
M.Theiss Hard- and Software (www.mtheiss.com/wcd.htm)
[3] European standard EN 673
[4] S.R. Wenham , M.A. Green and M.E. Watt, , "Applied Photovoltaics",
Appendix B, (Bridge Printery, Sydney, 1994).
[5] European standard EN 410
[6] M. List et al., Proc. 5 th ICCG (2004), p. 401, J. Puetz, A. Kurz, M.A. Aegerter
(eds.)
[7] S.J. Nadel et al., Proc. 4 th ICCG (2002), p. 53 , C.P. Klages, H. Gläser, M.A.
Aegerter (eds.)
[8] S.K.O'Leary, S.R.Johnson, P.K.Lim, J.Appl. Phys. Vol. 82, No. 7 (1997), p.
3334-3340
[9] P.Drude, Ann. Phys. 3 (1900), p. 369
[10] O. Treichel et al., Proc. 4 th ICCG (2002), p. 675, C.P. Klages, H. Gläser, M.A.
Aegerter (eds.)